IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 227-234

International Journal of Research in Information Technology (IJRIT)

www.ijrit.com

ISSN 2001-5569

Recovery of EMG Signals from the Mixture of ECG and EMG Signals 1

M. Nuthal Srinivasan, ME (Communication Systems) Student, Anna University of technology, Tiruchirapalli, BIT Campus, 1 [email protected],

2

T. Jaya Sankar Assistant Professor of ECE Anna University of technology, Tiruchirapalli, BIT Campus, 2 [email protected]

Abstract — Now a day in fast moving modern scientific world, medical diagnosis is one of the milestones for the human beings and animals. It is done by means of electronics and communication technology. Medical diagnosis is a cognitive process. A collection of useful data will make a diagnosis more effective in clinical process. One small error information or data result in wrong diagnosis. An effective technique for removing the presence of electrocardiograms (ECG) in surface electromyography (EMG) signals by means of time-variant harmonic modelling of the cardiac artefact. Heart rate and QRS complex variability, which often account for amplitude and frequency time variations of the ECG, are simultaneously captured by a set of third-order constantcoefficient polynomials modulating a stationary harmonic basis in the analysis window. Such a characterization allows us to significantly suppress ECG from the mixture by preserving most of the EMG signal content at low frequencies less than 20 Hz. Moreover, the resulting model is linear in parameters and the least-squares solution to the corresponding linear system of equations efficiently provides model parameter estimates. The comparative results suggest that the proposed method outperforms two reference methods in terms of the EMG preservation at low frequencies. Index Terms: ECG (Electrocardiogram), Harmonic modelling, Non stationary signals, Surface Electromyography (EMG). I. Introduction There are different methods to remove the ECG components from the EMG signal “Elimination of electrocardiogram contamination from electromyogram signals: An evaluation of currently used removal technique’’ .The simplest method consists of high-pass filtering EMG signal with a fourth order Butterworth filter at a cut-off frequency of 30Hz [1] .Surface Electromyography is an efficient technique for finding the activity of muscles. To extract the accurate information it is required to record a clean and undistorted electromyographic (EMG) signal [2]. The selection of an appropriate wavelet shapes and corresponding decision thresholding are major drawbacks from the users the main problem with “Changes in the action potential and contraction of isolated frog muscle after repetitive stimulation,” [3] – [7] method is that an important part of the EMG signals concerning the changes of negative after potentials is removed as well. It is known that the negative after potentials increase during fatigue and these changes could affect the amplitude of the EMG signal [1].

II. Proposed Method A Compact approach that addresses the issue of explicit nonstationary harmonic modelling of the ECG signal component. The motivation behind this approach arose from audio signal processing, where a similar scenario featuring a mixture of a quasi-harmonic signal component and a stochastic perturbation is often dealt. Herein, we model simultaneously both amplitude and frequency changes in the ECG signal component by means of a time-variant harmonic structure whose mean fundamental frequency is kept constant in the analysis window. It is shown that the time changes in an ECG harmonic are correctly captured by two constant-coefficients cubic polynomials each modulating a sine and a cosine function, respectively.[8] Electromyography (EMG) is a technique for evaluating and recording the electrical activity produced by skeletal muscles. EMG is performed using an instrument called an electromyograph, to produce a record called an electromyogram. An electromyograph detects the electrical potential generated by muscle cells when these cells are electrically or neurologically activated. The signals can be analysed to detect medical abnormalities, activation level, and recruitment order or to analyse the biomechanics of human or animal movement. EMG Signal consist of information regarding muscle activities and abnormalities, it has some unwanted information like ECG signal as noise.

M. Nuthal Srinivasan,IJRIT

227

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 227-234

ECG–EMG MIXTURE SIGNAL MODEL

ECG SIGNAL MODEL

FRQUENCY ESTIMATION

POLYNOMIAL COEFFICIENTS ESTIMATION

PERFORMANCE COMPARISON Fig.1.1: Block Diagram Electrocardiography (ECG or EKG from German: Elektrokardiogramm) is a transthoracic (across the thorax or chest) interpretation of the electrical activity of the heart over a period of time, as detected by electrodes attached to the outer surface of the skin and recorded by a device external to the body. The recording produced by this non-invasive procedure is termed as electrocardiogram (also ECG or EKG). An ECG test records the electrical activity of the heart. ECG is used to measure the rate and regularity of heartbeats, as well as the size and position of the chambers, the presence of any damage to the heart, and the effects of drugs or devices used to regulate

Fig.1.2: Raw EMG Signal the heart, such as a pacemaker. Most ECGs are performed for diagnostic or research purposes on human hearts, but may also be performed on animals, usually for diagnosis of heart abnormalities or research. Signal processing is a huge challenge since the actual signal value will be 0.5mV in an offset environment of 300mV. Other factors like AC power-supply interference, RF interference from surgery equipment, and implanted devices like pace makers and physiological monitoring systems can also impact accuracy. The main sources of noise in ECG are • • •

Baseline wander (low frequency noise) Power line interference ( 50Hz or 60Hz noise from power lines) Muscle noise (This noise is very difficult to remove as it is in the same region as the actual signal. It is usually corrected in software.)

M. Nuthal Srinivasan,IJRIT

228

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 227-234

Fig.1.3: ECG Signal Waveform

III.

Related Works

There are extensive research efforts dedicated to helping in removing the ECG noise from the raw EMG signal like Model-based filtering, compression and classification of the ECG, A dynamical model based on three coupled ordinary differential equations is introduced which is capable of generating realistic synthetic electrocardiogram (ECG) signals and also using wavelet independent decomposition analysis technique for removing the ECG signal from the EMG signal. The main drawbacks of the above research are important part of EMG signal concerning negative after potential is also removed during the ECG filtering. It is known that the negative after potentials increase during fatigue and these changes could affect the amplitude of the EMG signal significantly. In addition, it is found that these changes are reflected in the EMG spectrum within a frequency range below 10 Hz, Therefore, by filtering the EMG signal using a high-pass filter of 30Hz, valuable information of the EMG signal is removed when fatigue is analysed. [9]

IV.

ECG Signal Generation

Signal processing is a huge challenge since the actual signal value will be 0.5mV in an offset environment of 300mV. Other factors like AC power-supply interference, RF interference from surgery equipment, and implanted devices like pace makers and physiological monitoring systems can also impact accuracy. The main sources of noise in ECG are • Baseline wander (low frequency noise) • Power line interference (50Hz or 60Hz noise from power lines) • Muscle noise (This noise is very difficult to remove as it is in the same region as the actual signal. It is usually corrected in software.) The ECG signal can by generated by the principle of Fourier series. Any periodic functions which satisfy dirichlet’s condition can be expressed as a Series of scaled magnitudes of sine and cosine terms of frequencies which occur as multiples of fundamental frequency. The general Fourier series expression can be given as,[1]

  
     cos 

  sin 

2   1     
  

1      
  

 ,   1,2,3, ….  

M. Nuthal Srinivasan,IJRIT

1      
 $ 

 ,   1,2,3, ….  

229

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 227-234

It may notice that a single period of a ECG signal is a mixture of triangular and sinusoidal wave forms. Each significant feature of ECG signal can be represented by shifted and scaled versions one of these waveforms as shown below.
QRS, Q and S portions of ECG signal can be represented by triangular waveforms
P and T portions can be represented by triangular waveforms once we generate each of these portions, they can be added finally to get the ECG signal. Duration of P, Q, R, S, and T waves are
P-R interval 0.16s
S-T interval 0.18s
P interval 0.09s
QRS interval 0.11s The amplitude of the p wave is 0.25 mv. This p wave can be estimated by using this expression, % $⁄2 ∗  − 2) 2  ∗   ∗ cos 

$⁄2 ∗   2   +,  − 2    2 *


 

The amplitude of the q wave is 0.025 mv. This q wave can be obtained in the negative axis and is estimated by using this expression,[1]-[3] *


   cos  +,





    2⁄$ -  -  1 − cos 

cos 

  The amplitude of the qrs wave is 1.6 mv. This qrs wave can be estimated by using this expression, *

 
     cos 

2  +,

    2 −  

  2 ⁄$ -  -  1 − cos 





The amplitude of the s wave is 0.15 mv. This s wave can be obtained in the negative axis and is estimated by using this expression, *


   cos  +,





    2 ⁄$ -  -  1 − cos 

cos 

  The amplitude of the T wave is 0.35 mv. This T wave can be estimated by using this expression, % sin  ⁄2  − 2) $⁄2  2  − 2 
    2 +, 2  ∗    cos 

   *

V. Sine/Cosine Approximation The harmonic stationary f0 -basis modulated by the third-order time polynomials. Both amplitude and frequency time variations are compactly characterized by the polynomial coefficients. As a result, the above equation is linear in parameters, and

M. Nuthal Srinivasan,IJRIT

230

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 227-234

can be easily estimated by solving a linear system of equations. In order to check the Validity of the small-argument approximation, we have evaluated the sine/cosine approximation quality as a function of T in the following way: 6

./0 1  2 3 1$24
 1  5 3 124
 1 3+,

78 

∑ ∑ −  -

7:  Where

∑ ∑ − 1-

  $  ,

    ,   2
, 1- ,

1 are uniformly distributed time instants in the range [–T/2, T/2]. The error terms ;< and ;= are evaluated in decibels.

VI.

Polynomial Coefficients Estimation A

2 3 1  2> 1 >  ?@  ?, 1 − 24
, B@ 1 - − 24
, B, 1 A >+ A

3

3

3

3

3

5 3 1  5> 1 >  B@  B, 1 − 24
, ?@ 1 - − 24
, ?, 1 A >+

3

3

3

3

3

The coefficients 2 > 4 and 5 > 4 are efficiently estimated by means of the linear least-squares (LS) algorithm applied to above equation in the matrix form,   CD  7

Where D is the coefficient vector

D  %D, D- … D6 )

E 3 E

D6  %2 2, 2- 2A 5 5, 5- 5A ) 3

3

3

3

3

3

3

M is the signal model matrix which can be written as C  C8 C: C8 C: … … . . C8 C: ,

C8

6

M. Nuthal Srinivasan,IJRIT

,

-

-

6

6

E

$24
 1,  $24
 1-  … $24
 1I  H 1, $24
 1,  1- $24
 1-  … 1I $24
 1I  L  G 1, $24
 1,  1-- $24
 1-  … 1I- $24
 1I K A  1-A $24
 1-  … 1IA $24
 1I  F1, $24
 1, J

231

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 227-234

C:

6

24
 1,  24
 1-  … 24
 1I  H 1, 24
 1,  1- 24
 1-  … 1I 24
 1I  L  G 1, 24
 1,  1-- 24
 1-  … 1I- 24
 1I K 1 A 24
 1,  1-A 24
 1-  … 1IA 24
 1I  F, J

E

The vectors s and ε contain the signal samples and stochastic perturbation, respectively. The solution to the LS problem is a vector of the sought model parameters: DM  CN 

Where C N is the pseudo inverse matrix of M. the expression can be evaluated in many ways, among which we used the QR factorization of the over determined linear system M. once the parameters D are estimated, the EMG signal component is easily obtained as ̂.P0   − CDM For the estimation of EMG component from ECG-EMG mixture, the polynomial coefficients can be determined from the linear least square system of equations.

VII.

EXPERIMENTAL RESULTS

We collect some of the ECG signal and read them carefully about the characteristics after complete evaluation. We design ECG signal using time variant harmonic modelling using Fourier transform

WAVEFORMS 5 EMG-ECG mixture signal ESTIMATED EMG signal ECG signal

4

Fig 6.3: Estimated EMG Signal Amplitude

3

2

Fig 6.4: SINE/COSINE Approximation 1

VIII.

CONCLUSION

0

-1

0

5

10

15

20

25

Time(s)

SINE & COSINE APPROXIMATION 140 Sine approximation cosine approximation

Approximation quality [dB]

120

100

80

60

40

20

M. Nuthal Srinivasan,IJRIT

0.2

0.4

0.6

0.8

1 T[s]

1.2

1.4

1.6

1.8

232

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 227-234

EMG-to-Residual Ratio 14

EMG-to-Residual Ratio[dB]

12 10 8 6 4 2 0 -25

-20

-15 SNR[dB]

-10

-5

0

VIII. CONCLUSION We finally modelled ECG signal using this we can easily remove the ECG from the EMG Signal. The proposed method has been compared to two reference methods based on high-pass filtering and combined independent component analysis and wavelet transform, respectively. The experimental comparison results, regarding both artificial and real-world signals, show that in the analysis bandwidth 0–20 Hz, the proposed method outperforms the reference methods, as it introduces the smallest distortion in the EMG signal component.

IX. REFERENCES [1] Miroslav Zivanovic∗ and Miriam Gonz´alez-Izal, “Nonstationary Harmonic Modelling for ECG Removal in Surface EMG signals,” J. Biomedical., vol. 59, No. 6, June 2012 [2] C. J. De Luca, “The use of surface electromyography in biomechanics,” J. Appl. Biomech., vol. 13, no. 2, pp. 136–163, 1997. [3] J. D. Drake and J. P. Callaghan, “Elimination of electrocardiogram contamination from electromyogram signals: An evaluation of currently used removal techniques,” J. Electromyogr. Kinesiol., vol. 16, no. 2, pp. 175– 187, Apr. 2006. [4] M. S. Redfern, R. E. Hughes, and D. B. Chaffin, “High-pass filtering to remove electrocardiographic interference from torso EMG recordings,” Clin. Biomech., vol. 8, pp. 44–48, 1993. [5] J.Hanson and A. Persson, “Changes in the action potential and contraction of isolated frogmuscle after repetitive stimulation,” Acta. Physiol. Scand., vol. 81, pp. 340–348, 1971. [6] J. Hanson, “The effects of repetitive stimulation on the action potential and the twitch of rat muscle,” Acta. Physiol. Scand., vol. 90, pp. 387–400, 1974. [7] N. A. Dimitrova and G. V. Dimitrov, “Interpretation of EMG changes with fatigue: Facts, pitfalls, and fallacies,” J. Electromyogr. Kinesiol., vol. 13, no. 1, pp. 13–36, 2003. [8] Z. C. Lateva and K. C. McGill, “The physiological origin of the slow after wave in muscle action potentials,” Electroencephalogram. Clin. Neurophysiol., vol. 109, no. 5, pp. 462–469, Oct. 1998. [9] Y. Deng, W. Wolf, and R. Schnell, “New aspects to event-synchronous cancellation of ECG interference: An application of the method in diaphragmatic EMG signals,” IEEE Trans. Biomed. Eng., vol. 47, no. 9, pp. 1177–1184, Sep. 2000.

M. Nuthal Srinivasan,IJRIT

233

IJRIT International Journal of Research in Information Technology, Volume 2, Issue 5, May 2014, Pg: 227-234

[10] G. Lu, J. S. Brittain, P. Holland, J. Yianni, A. L. Green, J. F. Stein, T. Z. Aziz, and S. Wang, “Removing ECG noise from surface EMG signals using adaptive filtering,” Neurosci. Lett., vol. 462, no. 1, pp. 14–19, Oct. 2009.

M. Nuthal Srinivasan,IJRIT

234